Mysql查询计算跳出率

Question

我们需要计算有多少用户在我们的网站中弹跳。给定一个包含每个浏览量的GUID（用户ID），主机名和路径的表。如果每个主机名只有1个GUID，那么它被认为是反弹。

Table Pageviews: 
- GUID 
- Hostname 
- Path 

我能够做这个查询，但我认为它可能会得到改善。特别是在性能方面。

SELECT 1, (Bounces/All) * 100 AS `Bounce Rate` From (
SELECT 
    count(*) AS All, 
    (
     SELECT count(*) 
     FROM 
     (
      SELECT GUID 
      FROM pageviews 
      GROUP BY GUID 
      HAVING count(GUID) = 1 
     ) 
    ) AS Bounces 
    FROM pageviews 
) 

Answer 1

全球跳出率

组由GUID和主机名的数据（两次，用后者过滤，得到只反弹），然后外部联接在一起:)

SELECT count(bounces.guid) `bounces`, 
     count(uniqueUsers.guid) `total unique users`, 
     count(bounces.guid)/count(uniqueUsers.guid) * 100 `global bounce rate` 
FROM (
    SELECT guid, hostname 
    FROM PageViews 
    GROUP BY guid, hostname 
) uniqueUsers 
LEFT JOIN (
    SELECT guid, hostname 
    FROM PageViews 
    GROUP BY guid, hostname 
    HAVING COUNT(1) = 1 
) bounces 
ON uniqueUsers.guid = bounces.guid 
AND uniqueUsers.hostname = bounces.hostname 

实施例的结果：

bounces unique users global bounce rate 
------- ------------ ------------------ 
3   6    50.0000 

请注意，所有与'host1'对应的4'guid 3'命中仅计为1个唯一用户，但'gui d 1'同时击中了host1和host2，因此它统计了2个唯一用户（我认为这是所需的逻辑）。

---

每个主机跳出率

相同的，但与一组通过在外部查询:)

SELECT uniqueUsers.hostname, 
     count(bounces.guid) bounces, 
     count(uniqueUsers.guid) `unique users`, 
     count(bounces.guid)/count(uniqueUsers.guid) * 100 `global bounce rate` 
FROM (
    SELECT guid, hostname 
    FROM PageViews 
    GROUP BY guid, hostname 
) uniqueUsers 
LEFT JOIN (
    SELECT guid, hostname 
    FROM PageViews 
    GROUP BY guid, hostname 
    HAVING COUNT(1) = 1 
) bounces 
ON uniqueUsers.guid = bounces.guid 
AND uniqueUsers.hostname = bounces.hostname 
GROUP BY uniqueUsers.hostname; 

实施例的结果：

hostname bounces unique users bounce rate 
-------- ------- ------------ ----------- 
host1  2   4    50.0000 
host2  0   1    0.0000 
host3  1   1    100.0000 

---

示例数据

guid hostname path 
---- -------- ---- 
1  host1  irrelevant => bounce 1 
2  host1  irrelevant => bounce 2 
3  host1  irrelevant => non-bounce 1 (visit 1/4) 
3  host1  irrelevant 
3  host1  irrelevant 
3  host1  irrelevant 
4  host1  irrelevant => non-bounce 2 (visit 1/2) 
4  host1  irrelevant 
1  host2  irrelevant => non-bounce 3 (visit 1/2) 
1  host2  irrelevant 
2  host3  irrelevant => bounce 3